45 articles tagged with #policy-learning. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.
Researchers introduce DOM2, a diffusion-based offline multi-agent reinforcement learning algorithm that significantly improves policy expressiveness and generalization. The method achieves 20x better data efficiency and superior performance across standard benchmarks while maintaining robustness to environment shifts.
Researchers propose SOCD, an offline reinforcement learning algorithm that learns multi-user scheduling policies from pre-collected data without requiring real-time system interactions. The method combines diffusion models with critic guidance and Lagrangian optimization to handle delay-constrained resource allocation across applications like data centers and live streaming.
Researchers propose Bellman-Taylor score decoding, a novel deep reinforcement learning framework designed to handle Markov decision processes with state-dependent action constraints common in operations research. The method decouples policy learning into a Euclidean score space while maintaining feasibility through an action decoder, enabling standard DRL algorithms to optimize complex systems like queueing networks without architectural modifications.
Researchers introduce Bootstrapped Flow Q-Learning (BFQ), a new offline reinforcement learning method that achieves single-step action generation without multi-step denoising, improving computational efficiency and performance over existing diffusion-based approaches. The framework eliminates auxiliary networks and distillation procedures while maintaining high expressiveness, demonstrated through D4RL benchmark evaluations.
Researchers introduce AEGIS, a machine learning method that prevents robot manipulation failures by detecting high-risk steps and switching to a stronger policy only when needed. The system recovers 10.1% of failed trajectories while using stronger policies for just 38% of steps, demonstrating that selective escalation outperforms both blind backup policies and random triggering approaches.
Researchers propose Parameterized Diffusion Policy (PDP), a machine learning framework that enables diffusion models to learn controllable behaviors through low-dimensional parameters mapped to a semantic behavior manifold. This approach transforms diffusion models from stochastic noise generators into precise policy control tools, allowing smooth interpolation between strategies and adaptation to novel constraints without retraining.
Researchers propose DIBS, a decoupled behavioral cloning approach that improves reinforcement learning generalization by separating task-specific policy learning from evolution function learning. The method replaces noisy reward aggregation with stable supervision from teacher policies, achieving better training stability and zero-shot generalization compared to existing RL and meta-RL algorithms.
Researchers introduce Demo2Reward, a test-time optimization technique that improves Vision-Language Model (VLM) reward models by refining prompts based on a small number of expert demonstrations. The method reduces false positives in reward prediction without requiring additional model training, enabling more effective reinforcement learning in robotics applications including real-world scenarios.
SpeedAug is a new reinforcement learning framework that accelerates robotic policy execution by learning optimal task speeds rather than relying on conservative demonstration data. The method combines tempo-enriched policy learning with RL fine-tuning to achieve 1.8x faster real-world task throughput while maintaining success rates.
Researchers propose a learned 'rerooter' approach to improve Levin Tree Search for complex single-agent problems, eliminating the need for explicit subgoal generation. Three rerooter designs exploit state-space structure, learned heuristics, or hybrid signals to achieve scalable search with lower computational overhead and improved online training efficiency.
Researchers introduce World Action Verifier (WAV), a framework that enables world models to self-correct prediction errors by decomposing action-conditioned predictions into verifiable components: state plausibility and action reachability. The approach achieves 2x higher sample efficiency and 22% policy performance improvements across robotic control tasks by leveraging asymmetries in data availability and feature dimensionality.
AgensFlow is an open-source framework that treats multi-agent LLM coordination as a learnable policy problem rather than a fixed pipeline, enabling dynamic routing decisions across skill protocols, agent roles, and model bindings. Evaluated on distributed systems and security tasks, the framework demonstrates that learned coordination outperforms static designs while reducing exploration costs through warm-started policy graphs.
Researchers identify a critical flaw in robotic manipulation training: collecting diverse single-shot demonstrations paradoxically degrades performance due to estimation noise. Their proposed Anchor-Centric Adaptation (ACA) framework prioritizes repeated demonstrations at core tasks before expanding coverage, significantly improving robot reliability under strict data budgets.
Researchers introduced TAVIS, a comprehensive benchmark for evaluating active vision in imitation learning systems where robotic policies control their own gaze during manipulation tasks. The benchmark includes evaluation protocols, a novel metric (GALT) measuring anticipatory gaze, and baseline experiments showing that active vision benefits are task-dependent rather than universally beneficial.
Researchers introduce Aethelgard, an adaptive governance framework that addresses the capability overprovisioning problem in autonomous AI agents by dynamically restricting tool access based on task requirements. The system uses reinforcement learning to enforce least-privilege principles, reducing security exposure while maintaining operational efficiency.
Researchers introduce Hybrid-AIRL, an enhanced inverse reinforcement learning framework that combines adversarial learning with supervised expert guidance to improve reward function inference in complex, imperfect-information environments like poker. The method demonstrates superior sample efficiency and learning stability compared to traditional AIRL, particularly in settings with sparse and delayed rewards.
Researchers developed an improved Residual Reinforcement Learning method that uses uncertainty estimation to enhance sample efficiency and work with stochastic base policies. The approach outperformed existing methods in simulation benchmarks and demonstrated successful zero-shot sim-to-real transfer in real-world deployments.
Researchers developed an embodiment-aware transformer policy that improves cross-robot policy learning by injecting morphological information through kinematic tokens, topology-aware attention, and joint-attribute conditioning. This approach consistently outperforms baseline vision-language-action models across multiple robot embodiments.
Researchers propose BDGxRL, a novel framework using Diffusion Schrödinger Bridge to enable reinforcement learning agents to transfer policies across different domains without direct target environment access. The method aligns source domain transitions with target dynamics through offline demonstrations and introduces reward modulation for consistent learning.
The article title references learning policy representations in multiagent systems, which relates to AI research in multi-agent reinforcement learning. However, no article body content was provided for analysis.
